Ultrafast Medical Ultrasound imaging on a GPU
In conventional ultrasound imaging each image line is acquired sequentially. Since it takes a certain amount of time for the ultrasonic pulse to penetrate the tissue and to be reflected back, this conventional method poses physical limitations on the achievable frame-rate and resolution.
In ultrafast imaging, on the other hand, the entire field of view is acquired at once. While this potentially enables very high frame-rates (>1000fps), it also requires massive processing. Modern GPUs provide sufficient compute power to enable this new imaging paradigm.
While the main research in this area targets high-end desktop GPUs to be placed in the large and bulky standard ultrasound imaging systems, we want to explore more towards the application in mobile ultrasound systems such as the one currently developed in the UltrasoundToGo Nano-Tera Project.
Your task in this work is to implement an algorithm on the GPU that processes in a offline setup real captured ultrasound data as fast as possible.
As part of this project the student will:
- - Learn the fundamentals of ultrafast ultrasound imaging
- - Adapt and implement a given algorithm on the Tegra K1
- - Evaluate and compare the results (performance, power efficiency)
- 20% Theory, Algorithms and Simulation
- 80% Software Development
- Knowledge of Matlab and C/C++
- First experiences with GPU Programming (CUDA) are beneficial but not required